摘要：

针对互联网金融平台上的借贷违约行为,结合典型相关分析模型与复杂网络特征提取方法,建立违约风险量化模型,并利用ROC曲线与AUC值对模型效果进行评价。将该模型应用于某互联网借贷平台实际交易数据中,并与决策树模型进行比较分析,结果表明,该模型能够有效反映用户违约特征与复杂网络特征的关联效应,其AUC值能达到0.85左右,且具有更强的稳定性与鲁棒性。

关键词: 互联网金融, 典型相关分析, 复杂网络, 中心性, 风险量化

Abstract:

Aiming at the loan default behavior on the Internet financial platform,combining the canonical correlation analysis model and the complex network feature extraction method,the quantitative model of default risk is established,and the ROC curve and AUC value are used to evaluate the effect of the model.The model is applied to the actual transaction data of an Internet lending platform and compared with the decision tree model.Results show that the proposed model can effectively reflect the correlation effect between user default features and complex network features,the AUC value of the model can reach about 0.85,and it has stronger stability and robustness.

Key words: Internet finance, canonical correlation analysis, complex network, centrality, risk quantification

中图分类号: 

• TP391